Asymptotically compatible meshfree discretization of state-based peridynamics for linearly elastic composite materials

Nathaniel Trask, Benjamin Huntington, David J. LittlewoodarXiv 2019 · OpenReview

Introduction

Peridynamics has emerged as a powerful nonlocal alternative to classical continuum mechanics, particularly for modeling material failure and fracture. Unlike standard Finite Element Analysis (FEA), which relies on partial differential equations that become ill-defined at discontinuities, peridynamics utilizes integral equations. This allows for a natural representation of cracks and material separation. However, the transition from continuous nonlocal models to discrete numerical implementations has historically been plagued by "surface effects" and a lack of asymptotic compatibility (AC)—the property where a numerical solution correctly recovers the classical local solution as both the discretization size ($h$) and the nonlocal interaction radius ($\delta$) vanish.

This paper presents a novel meshfree discretization for state-based peridynamics that achieves asymptotic compatibility for linearly elastic composite materials. By integrating an optimization-based quadrature (OBQ) framework with a specialized correction for nonlocal dilatation, the authors provide a robust solution to the consistency and convergence issues that have limited the practical application of peridynamics.

The Challenge: Surface Effects and Convergence

The primary mathematical hurdle in nonlocal mechanics is ensuring that the discrete model behaves predictably as the resolution increases. In many meshfree peridynamic simulations, points near the boundary of a domain or near a crack face experience a "truncation" of their interaction neighborhood. This leads to spurious errors known as surface effects, which prevent the model from recovering the correct traction-free boundary conditions of classical elasticity. Furthermore, standard one-point quadrature schemes often fail to accurately integrate the singular or near-singular kernels characteristic of peridynamic models, leading to poor convergence rates.

Methodology: OBQ and Dilatation Correction

The authors address these challenges through a two-pronged numerical strategy:

  1. Optimization-Based Quadrature (OBQ): Rather than relying on simplistic integration rules, the authors employ an OBQ framework. For every point in the discretization, a small optimization problem is solved to determine quadrature weights that exactly integrate a set of basis polynomials. This ensures that the discrete operators maintain the mathematical properties of their continuous counterparts, effectively fixing integration errors that are often mistaken for physical modeling errors.
  2. Modified Dilatation with a Correction Tensor: State-based peridynamics relies heavily on "dilatation" (a measure of local volume change). To correct for neighborhood truncation near boundaries, the authors introduce a correction tensor $M(x)$. Inspired by techniques in Smoothed Particle Hydrodynamics (SPH), this tensor re-normalizes the dilatation calculation. This allows the model to recover the local divergence operator even at free surfaces, ensuring that the physics remains consistent across the entire domain.
  3. Bond-Breaking for Fracture: The framework maintains the standard peridynamic approach to fracture, where bonds between points are "broken" once they exceed a critical strain threshold. The OBQ and dilatation corrections are shown to remain effective even as the topology of the material changes during crack propagation.

Numerical Verification and Results

The efficacy of the proposed method was validated through a series of rigorous analytic benchmarks rather than experimental data, focusing on mathematical convergence:

  • Quadratic Patch Tests: The method successfully reproduced simple polynomial displacement fields, confirming its basic consistency.
  • Manufactured Solutions: Using smooth trigonometric functions, the authors demonstrated $O(h^2)$ second-order convergence in the interior of the domain.
  • Michell Solution (Hole in a Plate): This classic elasticity problem tested the method's ability to handle curvilinear free surfaces. The method achieved $O(h)$ convergence, the theoretical optimum for such geometries.
  • Cylindrical Inclusion: To test composite material performance, the authors modeled a stiff inclusion within a softer matrix. The method remained stable even with shear modulus ratios as high as 256:1.

A significant finding was the method's performance in the near-incompressible limit. While many numerical methods suffer from "locking" as the Poisson ratio $\nu$ approaches 0.5, this discretization remained stable and convergent even at $\nu = 0.495$.

Conclusion and Future Outlook

The research demonstrates that many of the historical difficulties in peridynamics—particularly surface effects—stem from numerical integration errors rather than inherent flaws in the nonlocal theory. By employing OBQ and the $M(x)$ correction tensor, the authors have created a framework that is both asymptotically compatible and robust for complex composite geometries.

While the current work is limited to linear elasticity and static loading, it provides a foundation for future extensions. The authors intend to apply this framework to nonlinear elastoplastic mechanics and ductile fracture, where the ability to handle large deformations and evolving discontinuities is paramount. Future work will also address the observed grid sensitivity in composite materials, where non-uniform grids can lead to convergence stagnation at high precision.

Summary updated

Abstract

State-based peridynamic models provide an important extension of bond-based models that allow the description of general linearly elastic materials. Meshfree discretizations of these nonlocal models are attractive due to their ability to naturally handle fracture. However, singularities in the integral operators have historically proven problematic when seeking convergent discretizations. We utilize a recently introduced optimization-based quadrature framework to obtain an asymptotically compatible scheme able to discretely recover local linear elasticity as the nonlocal interaction is reduced at the same rate as the grid spacing. By introducing a correction to the definition of nonlocal dilitation, surface effects for problems involving bond-breaking and free surfaces are avoided without the need to modify the material model. We use a series of analytic benchmarks to validate the consistency of this approach, illustrating second-order convergence for the Dirichlet problem, and first-order convergence for problems involving curvilinear free surfaces and composite materials. We additionally illustrate that these results hold for material parameters chosen in the near-incompressible limit.
BibTeX
@misc{trask2019asymptotically,
  author = {Nathaniel Trask and Benjamin Huntington and David J. Littlewood},
  title = {{Asymptotically compatible meshfree discretization of state-based peridynamics for linearly elastic composite materials}},
  year = {2019},
  eprint = {1903.00383},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1903.00383},
}
Figure